Rapid model prediction and history matching using scientific machine learning for geologic carbon storage at the Illinois Basin-Decatur Project site
Explore the source record for details and available documents.
Engineering topics
Publications and source records attributed to Lee, Jonghyun.
Explore the source record for details and available documents.
Estimating spatially distributed properties such as permeability from available sparse measurements is a great challenge in efficient subsurface CO 2 storage operations. In this paper, a deep generative model that can accurately capture complex subsurface structure is tested with an ensemble-based inversion method for accurate and accelerated characterization of CO 2 storage sites. We chose Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) for its realistic reservoir property representation and Ensemble Smoother with Multiple Data Assimilation (ES-MDA) for its robust data fitting and uncertainty quantification capability. WGAN-GP are trained to generate high-dimensional permeability fields from a low-dimensional latent space and ES-MDA then updates the latent variables by assimilating available measurements. Several subsurface site characterization examples including Gaussian, channelized, and fractured reservoirs are used to evaluate the accuracy and computational efficiency of the proposed method and the main features of the unknown permeability fields are characterized accurately with reliable uncertainty quantification. Furthermore, the estimation performance is compared with a widely-used variational, i.e., optimization-based, inversion approach, and the proposed approach outperforms the variational inversion method in several benchmark cases. We explain such superior performance by visualizing the objective function in the latent space: because of nonlinear and aggressive dimension reduction via generative modeling, the objective function surface becomes extremely complex while the ensemble approximation can smooth out the multi-modal surface during the minimization. This suggests that the ensemble-based approach works well over the variational approach when combined with deep generative models at the cost of forward model runs unless convergence-ensuring modifications are implemented in the variational inversion.
Explore the source record for details and available documents.
Metals and metallic alloys often have high melting temperatures and highly reactive liquids. Processing reactive liquids in containers can result in significant contamination and limited undercooling. This is particularly true for molten silicon and it alloys. Silicon is commonly termed "the universal solvent". The viscosity, surface tension, and density of several silicon‐transition metal alloys were determined using the Electrostatic Levitator system at the Marshall Space Flight Center. The temperature dependence of the viscosity followed an Arrhenius dependence, and the surface tension followed a linear temperature dependence. The density of the melts, including the undercooled region, showed a linear behavior as well. Viscosity and surface tension values were obtain for several of the alloys in the undercooled region.
During the past years, extensive collaborative research has been done to understand phase selection in undercooled metals using novel containerless processing techniques such as electrostatic and electromagnetic levitation. Of major interest is controlling a two-step solidification process, double recalescence, in which the metastable phase forms first and then transforms to the stable phase after a certain delay time. The previous research has shown that the delay time is greatly influenced by the internal convection velocity. In the prediction of internal flow, the fidelity of the results depends on the accuracy of the material properties. This research focuses on the measurements of density and viscosity of Fe-Co alloys which will be used for the fluid simulations whose results will support upcoming International Space Station flight experiments.
A non-contact technique for measuring the creep properties of materials has been developed and validated as part of a collaboration among the University of Massachusetts, NASA Marshall Space Flight Center Electrostatic Levitation Facility (ESL), and the University of Tennessee. This novel method has several advantages over conventional creep testing. The sample is deformed by the centripetal acceleration from the rapid rotation, and the deformed shapes are analyzed to determine the strain. Since there is no contact with grips, there is no theoretical maximum temperature and no concern about chemical compatibility. Materials may be tested at the service temperature even for extreme environments such as rocket nozzles, or above the service temperature for accelerated testing of materials for applications such as jet engines or turbopumps for liquid-fueled engines. The creep measurements have been demonstrated to 2400 C with niobium, while the test facility, the NASA MSFC ESL, has processed materials up to 3400 C. Furthermore, the ESL creep method employs a distribution of stress to determine the stress exponent from a single test, versus the many tests required by conventional methods. Determination of the stress exponent from the ESL creep tests requires very precise measurement of the surface shape of the deformed sample for comparison to deformations predicted by finite element models for different stress exponents. An error analysis shows that the stress exponent can be determined to about 1% accuracy with the current methods and apparatus. The creep properties of single-crystal niobium at 1985 C showed excellent agreement with conventional tests performed according to ASTM Standard E-139. Tests on other metals, ceramics, and composites relevant to rocket propulsion and turbine engines are underway.
State-of-the-art technologies for hypersonic aircraft, nuclear electric/thermal propulsion for spacecraft, and more efficient jet engines are driving ever more demanding needs for high-temperature (>2000 C) materials. At such high temperatures, creep rises as one of the most important design factors to be considered. Since conventional measurement techniques for creep resistance are limited to about 17OO0C, a new technique is in demand for higher temperatures. This paper presents a non-contact method using electrostatic levitation (ESL) which is applicable to both metallic and non-metallic materials. The samples were rotated quickly enough to cause creep deformation by centrifugal acceleration. The deformation of the samples was captured with a high speed camera and then the images were analyzed to estimate creep resistance. Finite element analyses were performed and compared to the experiments to verify the new method. Results are presented for niobium and tungsten, representative refractory materials at 2300 C and 2700 C respectively.
Responsive access to space requires re-use of components such as rocket nozzles that operate at extremely high temperatures. For such applications, new ultra-hightemperature materials that can operate over 2,000 C are required. At the temperatures higher than the fifty percent of the melting temperature, the characterization of creep properties is indispensable. Since conventional methods for the measurement of creep is limited below 1,700 C, a new technique that can be applied at higher temperatures is strongly demanded. This research develops a non-contact method for the measurement of creep at the temperatures over 2,300 C. Using the electrostatic levitator in NASA MSFC, a spherical sample was rotated to cause creep deformation by centrifugal acceleration. The deforming sample was captured with a digital camera and analyzed to measure creep deformation. Numerical and analytical analyses have also been conducted to compare the experimental results. Analytical, numerical, and experimental results showed a good agreement with one another.